LingoDB is a new analytical database system that blurs the lines between databases and compilers.

We provide a Dockerfile that can be used to build and experiment with LingoDB. For a simple build, just execute this command in the repository. Note that no submodules need to be checked out.

make build-docker

If you just want to see results for the TPC-H queries just run this command. Note that docker comes with a slight runtime overhead (passing --privileged helps a bit).

make reproduce

1. Ensure you have a machine with sufficient compute power and space (16 GiB RAM, >16 GiB disk space, preferably many cores).
2. Initialize submodules: git submodule update --init --recursive
3. Install (system) dependencies:
   • Ubuntu: libjemalloc-dev libboost-dev libboost-filesystem-dev libboost-system-dev libboost-regex-dev python-dev autoconf flex bison
   • Fedora: autoconf cmake ninja jemalloc-devel boost-devel boost-filesystem boost-system python3-devel flex bison lcov
4. Setup a Python virtual environment and install dependencies:

   python3 -m venv venv
   . venv/bin/activate
   pip install -r torch-mlir/requirements.txt
   pip install requests Cython moz_sql_parser numpy pandas pyarrow

5. Build dependencies: make dependencies (this builds LLVM/MLIR/Clang, Arrow and PyArrow)
6. Build & test: make run-test (this compiles in debug mode; incremental builds are supported)
7. Run TPC-H (SF=1) benchmarks: make run-benchmark (this generates TPC-H data and compiles in release mode; incremental builds are supported)